European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331

Contents lists available at ScienceDirect

European Journal of Pharmaceutics and Biopharmaceutics

journal homepage: www.elsevier .com/locate /e jpb

Research paper

Complex coacervates of hyaluronic acid and lysozyme: Effect on proteinstructure and physical stability

http://dx.doi.org/10.1016/j.ejpb.2014.09.0010939-6411/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Department of Pharmacy, Faculty of Health and MedicalSciences, University of Copenhagen, Universitetsparken 2, 2100 Copenhagen O,Denmark. Tel.: +45 35 33 60 57.

E-mail address: jorrit.water@sund.ku.dk (J.J. Water).

Jorrit J. Water a,⇑, Malthe M. Schack a,d, Adrian Velazquez-Campoy b,c, Morten J. Maltesen d,Marco van de Weert a, Lene Jorgensen a

a Department of Pharmacy, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen O, Denmarkb Institute of Biocomputation and Physics of Complex Systems (BIFI), Joint-Unit IQFR-CSIC-BIFI, Department of Biochemistry and Molecular and Cell Biology, University of Zaragoza,Zaragoza, Spainc Fundacion ARAID, Government of Aragon, Zaragoza, Spaind Novozymes A/S, Biopharma Application Development, Bagsvaerd, Denmark

a r t i c l e i n f o

Article history:Received 29 April 2014Accepted in revised form 1 September 2014Available online 16 September 2014

Keywords:Isothermal titration calorimetryDifferential scanning calorimetryStabilityBinding characteristicsDifferential scanning fluorimetryHydrogelBovine serum albuminComplex coacervationHyaluronanSodium hyaluronate

a b s t r a c t

Complex coacervates of hyaluronic acid and lysozyme, a model protein, were formed by ionic interactionusing bulk mixing and were characterized in terms of binding stoichiometry and protein structure andstability. The complexes were formed at pH 7.2 at low ionic strength (6 mM) and the binding stoichiom-etry was determined using solution depletion and isothermal titration calorimetry. The binding stoichi-ometry of lysozyme to hyaluronic acid (870 kDa) determined by solution depletion was found to be225.9 ± 6.6 mol, or 0.1 bound lysozyme molecules per hyaluronic acid monomer. This corresponded wellwith that obtained by isothermal titration calorimetry of 0.09 bound lysozyme molecules per hyaluronicacid monomer. The complexation did not alter the secondary structure of lysozyme measured by Fourier-transform infrared spectroscopy overlap analysis and had no significant impact on the Tm of lysozymedetermined by differential scanning calorimetry. Furthermore, the protein stability of lysozyme wasfound to be improved upon complexation during a 12-weeks storage study at room temperature, asshown by a significant increase in recovered protein when complexed (94 ± 2% and 102 ± 5% dependingon the polymer-protein weight to weight ratio) compared to 89 ± 2% recovery for uncomplexed protein.This study shows the potential of hyaluronic acid to be used in combination with complex coacervationto increase the physical stability of pharmaceutical protein formulations.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

Complex coacervation is the binding of two oppositely chargedmolecules driven mainly by electrostatic interactions that resultsin a liquid–liquid phase separation between the bulk solutionand the soluble complexes [6]. In the majority of the cases thesemolecules are macromolecules, e.g. polyelectrolytes such as, pro-teins, nucleic acids, dendrimers or linear polymers. The coacervatesare in a soluble state, in contrast to complex flocculates, which areformed upon precipitation of the interacting macromolecules inthe case of highly efficient interactions, thereby expelling thewater from the complex structure. In many cases, this solublenature of complex coacervates leaves the biofunctionality of the

complexed macromolecules intact, thereby making them interest-ing for formulation of pharmaceutical proteins [14]. A wide varietyof complex coacervates systems have been reported and investi-gated in the literature, e.g. protein–protein [5,32], protein–polymer[13,31], nucleic acid–polymer [16,26,24] and nucleic acid–dendrimer[11,19]. A more complete overview can be found in recent reviewpapers [14,30].

The physicochemical properties of complex coacervates makethem interesting as potential drug delivery systems, mainly dueto their earlier mentioned ability to retain the biofunctionalityof a pharmaceutically relevant macromolecule. Different drugdelivery strategies have been explored by exploiting the physico-chemical nature of complex coacervates such as using them asstabilizing agents to preserve biofunctionality [35], reduce enzy-matic cleavage as well as physical stability of the biological byinhibition of aggregation [7] or denaturation [23], or for entrap-ment of the molecule in complexes used for controlled releasedelivery. However, the effect is also dependent on the complexed

326 J.J. Water et al. / European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331

macromolecule e.g. the physical stability of lysozyme (LZ) wasreduced upon complexation with heparin [31].

Hyaluronic acid (HA) is a linear anionic polysaccharide consist-ing of N-acetyl-D-glucosamine and D-glucuronic acid disaccharideunits bound by alternating b-(1 �> 3) b-(1 �> 4) glucosidic bondsfirst discovered by Meyer and Palmer [21]. It is mainly used forbasic clinical applications like viscoaugmentation and viscoprotec-tion [4,34] due to its high viscosity caused by the arrangement ofthe molecules in structured networks based upon hydrophobicpatches and intermolecular hydrogen bonds [28]. HA is biocompat-ible and biodegradable, and less likely to cause undesired immuneresponses due to its presence in almost all biological fluids and tis-sues of vertebrates, making it a promising polymer for drug deliv-ery applications.

The complex coacervation behavior of HA has been investigatedwith different polyelectrolyte polymers such as chitosan [17] andsilk fibroin [20]. Other formulations of HA for drug delivery appli-cations include complex formation in the presence of cationic poly-electrolytes for delivery of nucleic acids [25,1], and covalentlybound or cross-linked HA molecules forming hydrogel networksand depots for controlled release [8,12].

This work investigated the use of HA as a complex coacervateforming polyelectrolyte by binding positively charged proteinsand thereby potentially increase their physical stability. Lysozymewas chosen as model protein, as it has been shown earlier that itsstability is decreased upon complex coacervation with heparin[31]. BSA was also investigated, as it is mainly negatively chargedat physiological pH and did not form macroscopic complex coacer-vates upon mixing with HA under the studied conditions. However,coacervation has been reported in literature by Du et al. [10] underdifferent conditions. The binding stoichiometry of the HA–LZ com-plexes at low ionic strength conditions was investigated and theinfluence of the complexation behavior on the stability of the pro-tein was determined. The thermal unfolding and structure of LZand BSA was studied with differential scanning calorimetry(DSC), Fourier-transform infrared spectroscopy (FTIR), differentialscanning fluorimetry (DSF) and the stoichiometry determined byisothermal titration calorimetry (ITC) and solution depletion.

2. Material and methods

2.1. Materials

Hyaluronic acid with an average molecular weight of 870 kDawas supplied by Novozymes A/S (Bagsvaerd, Denmark). Hen eggwhite lysozyme and bovine serum albumin (>98%, fraction V) wereobtained from Sigma (St. Louis, US). HEPES buffer salt was pur-chased from Applichem (Darmstadt, Germany), sodium chloride(NaCl) (Sigma, St. Louis, US) and Milli-Q water (Millipore, Billerica,US) was used in all buffers and sample preparations.

2.2. Buffer and sample preparation

HEPES buffer (25 mM) pH 7.2 was prepared with an ionicstrength of 6 mM. The ionic strength was adjusted by addition ofNaCl and all buffers were filtered (0.22 lm filter) before use. Disso-lution of HA and LZ was performed overnight at room temperature.HA solutions were continuously stirred to speed up the dissolutionprocess. Before each experiment the concentration of LZ(E1%,280 = 26.4) [3] was determined using a NanoDrop 2000c(Thermo Fisher Scientific, Waltham, US) and adjusted to therequired concentration, and if necessary samples were filtered(0.22 lm) shortly before adjusting the concentration. When pre-paring complex coacervates, samples were allowed to incubate at

least 1 h to ensure complete complexation before the differentexperiments.

2.3. Differential scanning calorimetry

DSC experiments were conducted on a MicroCal VP-DSC system(GE Healthcare, Little Chalfont, UK). All samples were prepared inHEPES buffer (pH 7.2, ionic strength = 6 mM) at a concentrationof 2.4 mg/mL LZ and 0.768 mg/mL HA. Samples and buffers weredegassed for 5 min by stirring under vacuum before loading intothe sample cell at a 1:0.32 (w/w) ratio LZ:HA and incubated for1 h before measurements were conducted. Thermograms wereobtained by scanning at 90 �C/h from 20 �C to 100 �C and meltingtemperatures (Tm) were determined using the Origin 7.0 softwarepackage (OriginLab, Northampton, US).

2.4. Solution depletion

To determine the binding stoichiometry of LZ and HA solutiondepletion experiments were performed. The concentrations of LZand BSA were assumed to be well above the dissociation constantfor the most important binding sites on the HA. The average num-ber of LZ molecules per HA chain can then be determined using thefollowing formula as proposed by [31]:

½LZ�free½LZ�initial

¼ 1� n� ½HA�initial½LZ�initial

where [LZ]free is the free LZ concentration measured in the super-natant, and [LZ]initial and [HA]initial are the starting concentra-tions of LZ and HA in solution. The binding stoichiometry (n) isdetermined as the average number of LZ molecules bound per HAchain.

HA dilution series were prepared by serial 2-fold dilution(0.768 mg/mL to 0.144 mg/mL) of the stock solutions in the differ-ent ionic strength buffers. The HA solutions were then added to afixed amount of LZ (2.6 mg/mL) in a 1:5 (v/v) ratio and incubatedfor 1 h at room temperature. After incubation the samples werespun down in a tabletop centrifuge at 18,000g for 15 min. The con-centration of free LZ in the supernatant of the samples was thendetermined by UV-absorbance at 280 nm using the NanoDrop2000c and binding stoichiometry was calculated.

2.5. Fourier-transform infrared spectroscopy

Infrared spectra were recorded with a Bomen MB-100 seriesFTIR spectrometer (Quebec, Canada). For each experiment 15 lLof sample was loaded onto a calcium fluoride (CaF2) crystalwindow with a 6 lm path length (Biotools, Wauconda, US). TheIR-spectra were collected in transmission mode (single beam)averaging 256 scans with a 2 cm�1 resolution at room temperature.

The spectra were analyzed according to a standardized protocolas described previously by [9] using the GRAMS AI spectroscopysoftware package (Thermo Scientific, Waltham, US). First the spec-tra of the buffer solution and water vapor were subtracted from thesample spectra. The second derivatives of the resulting spectrawere obtained with an eleven or fifteen point smoothingSavitzky-Golay derivative function (spectral window of 22 or30 cm�1, respectively). The spectra were subsequently baselinecorrected and area-normalized (amide I region, 1595–1705 cm�1)and finally the area overlap method was used [15] to determinethe area overlap.

The complexes contained a final ratio of respectively 1:0.32w/w LZ:HA to ensure complete complexation of all LZ in solution,based on solution depletion data. Spectra of free LZ were obtained

0.000 0.001 0.002 0.003 0.004 0.005 0.0060.0

0.2

0.4

0.6

0.8

1.0

[HA]ini/[LZ]ini (mol)

[LZ]

free

/[LZ]

ini

Fig. 1. Solution depletion data for different ionic strengths, circles 6 mM andsquares 100 mM (mean ± range, n = 2).

J.J. Water et al. / European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331 327

at a concentration of 40 mg/mL. The samples were loaded by trans-ferring the complexes on to the CaF2 crystal window.

2.6. Isothermal titration calorimetry

ITC experiments were performed on a TA Instruments Nano-ITCsystem (New Castle, US). All stock solutions of LZ and HA were pre-pared in the same buffer (ionic strength = 6 mM) to avoid largeheats of mixing. The analyte solutions for the HA titrations intoLZ (0.75 mg/mL) solutions were filtered (0.22 lM) and degassedwhile stirring for 5–10 min before being loaded into the titrationchamber. The HA titrant solution (2.64 mg/mL) was degassed whilestirring 5–10 min before being loaded into the titration syringe.The titration experiments consisted of 20 injections of 10 lL whilecontinuously stirring at 250 rpm at 25 �C starting with a low vol-ume (3.15 lL) initial injection.

The experimental data were analyzed using the McGhee-vonHippel model for the cooperative binding of molecules (LZ) to a lin-ear macromolecule (HA) [33]. Given that HA is injected into LZ, theexperiments correspond to reverse titrations [33]. The averagemolecular weight of HA is 870 kDa and the molecular weight of adisaccharide monomer is 401 Da, from which an average numberof 2170 monomers per HA molecule can be estimated. Stoichiom-etry and binding parameters can be estimated through non-linearregression of the titration data [33], in particular, the number ofmonomers occupied by the bound ligand (L, ligand size in macro-molecule subunits), the dissociation constant (Kd), the bindingenthalpy per mole of bound ligand (DH), the cooperativity interac-tion constant (x), and the cooperativity enthalpy (Dh).

2.7. Differential scanning fluorimetry

DSF experiments were performed on a Stratagene AgilentMx3005p QPCR instrument (Santa Clara, US) fitted with custom fil-ter sets. 96-well non-skirted, regular profile, opaque white qPCRplates were used and closed with optical sealing tape. The finalprotein concentration in each well was 2.4 mg/mL (LZ) or0.04 mg/mL (BSA) in HEPES buffer (pH 7.2, ionic strength = 6 mM).Samples were incubated 1 h at room temperature before transfer-ring to the 96-well plate. Two different HA concentrations (0 or0.768 mg/mL) were tested in buffer at three different ionicstrengths adjusted with sodium chloride concentrations (0,50 mM and 100 mM) for each protein.

SYPRO orange stock solutions were made just before measure-ments in black test tubes and 0.5 lL of the stock solution wasadded to each well and mixed carefully resulting in a final samplevolume of 50 lL. The final concentration of SYPRO orange was 5Xfor LZ and 2.5X for BSA. The plates were centrifuged at 100 g forone min before measurements. All measurements were performedin triplicate with an excitation wavelength of 492 nm and emissionmeasured at 610 nm. The samples were heated stepwise from 25to 95 �C at 1 �C per min. Each temperature rise was followed by30 s equilibration, with three consecutive measurements at theend of each cycle. The melting temperatures (Tm) were calculatedbased on the first derivative of the fluorescence intensity.

2.8. Stability study

Samples were prepared by addition of HA at two different finalconcentrations (0.35 mg/mL or 0.70 mg/mL) to either LZ or BSA(2.4 mg/mL final concentration) in low ionic strength conditions(I = 6 mM). Control samples were prepared at high ionic strength(I = 100 mM) to inhibit complex coacervation. The samples weresubsequently stored at room temperature for 12 weeks andafterward kept at �20 �C until further analysis. For analysis thecomplexes were disassociated by addition of sodium chloride to

a final concentration of 0.4 M. The samples were analyzed byRP-HPLC on a Shimadzu RP-HPLC (Shimadzu Corporation, Kyoto,Japan) with a C18 column (50 mm � 4.6 mm, 2.6 lm, Kinetex,Phenomenex, Allerød, Denmark) at a constant flow of 1.85 mL/min; the protein was detected using UV absorbance at 218 nm.The mobile phase consisted of solvent A (5:95 Acetonitrile:MilliQand 0.1% TFA(v/v)) and solvent B (95:5 Acetonitrile:MilliQ and0.1%TFA (v/v)). LZ samples were eluted using a gradient from15–55% solvent B over 7.5 min, whereas BSA samples were elutedusing 5–75% solvent B over 7.5 min. The amount of recovered pro-tein was expressed as the reduction (in percentage) between theAUC of the stored and the control sample (freshly prepared).

2.9. Statistical analysis

Experiments were performed in triplicates unless indicatedotherwise and values were given as means ± standard deviation(SD). The FTIR area overlap data were analyzed using a one-wayANOVA at a 0.05 significance level and the DSF data were analyzedusing a two-way ANOVA at a 0.05 significance level. All statisticalanalysis was carried out using GraphPad Prism 6 (La Jolla, CA, US)software.

3. Results

3.1. Complex characterization

In order to characterize the complexation behavior of HA and LZsolution depletion experiments were performed. To obtain satura-tion of the HA binding sites at low concentrations of HA a highconcentration of LZ (2.4 mg/mL, 164.4 lM) was added. Under theseconditions transparent complex coacervates were observed uponmixing of the two compounds. Fig. 1 shows the ratio between freeLZ in the supernatant and the initial concentration of free LZ vs. theHA:LZ ratio, at high (100 mM) and low (6 mM) ionic strengths.These data show a linear correlation between the free LZ concen-tration in the supernatant and the amount of HA added at low ionicstrength (6 mM). The solution was depleted of LZ at a molar ratio of0.0046 ± 0.001 HA:LZ and above these ratios no free LZ wasdetected in the supernatant. The binding stoichiometry was deter-mined to be 225.9 ± 6.6, meaning that 225.9 ± 6.6 mol of LZ can bebound by one mole of HA at an ionic strength of 6 mM, which cor-responds to a binding of approximately 0.1 LZ molecules per HAmonomer. At high ionic strength (100 mM) the complex coacer-vates formation is completely inhibited due the presence of coun-ter ions resulting in a binding stoichiometry of n = 0. Samplescontaining BSA instead of LZ were also prepared for solution

328 J.J. Water et al. / European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331

depletion experiments and as expected showed no visible complexcoacervation upon mixing in any ratio as well as no reduction ofthe BSA concentration measured in the supernatant (data notshown), indicating that no macroscopic complex coacervates areformed between the negatively charged polymer and protein. Notethat this does not exclude the formation of soluble complexeswhich could be present as indicated by Lenormand et al. [18]and the work of Anderot et al. [2], which shows interactionsbetween HSA and heparin, even though both molecules carry thesame net anionic charge density.

Titration experiments by ITC were conducted in order to furthercharacterize the interaction between HA and LZ. From these exper-iments (Fig. 2), stoichiometry and binding parameters could be cal-culated for the HA/LZ interaction. The estimated value of L(number of monomers occupied by the bound ligand) was11.3 ± 0.2, showing that each LZ molecule bound to HA occupiedan effective size of 11.3 disaccharide subunits. This correspondsto 0.09 bound molecules of LZ per HA monomer, in close agree-ment to the results obtained by solution depletion experiments(0.1 LZ molecules per HA monomer). The dissociation constant,Kd, for the interaction of each LZ molecule bound to HA was13 ± 1 lM, which corresponds to a binding Gibbs energy of�6.8 ± 0.1 kcal/mol, and confirms the assumptions made for thesolution depletion experiment with regards to saturation of theHA binding sites by addition of high concentrations of LZ. The bind-ing of LZ to HA was entropically driven, characterized by a slightlyfavorable binding enthalpy (DH) of �0.3 ± 0.2 kcal/mol and a highlyfavorable entropic contribution (�TDS) of �6.5 ± 0.2 kcal/mol.In addition, the binding of LZ molecules to HA showed significantpositive cooperativity, with a cooperativity interaction constant(x) of 38 ± 4, which corresponds to a cooperativity Gibbs energyof �2.2 ± 0.1 kcal/mol, and a cooperativity enthalpy (Dh) of�0.8 ± 0.2 kcal/mol.

0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014-600

-500

-400

-300

-200

-100

0

100

-0.1

0.0

0.1

0.2

0.3

0.4

0 30 60 90 120 150 180 210time (min)

dQ/d

t (µc

al/s

)

[HA]T/[LZ]T

Q (k

cal/m

ol o

f inj

ecta

nt)

Fig. 2. Reversed calorimetric titration of HA into LZ (n = 3). (Upper panel)Thermogram showing the thermal power as a function of time; (lower panel)binding isotherm showing the normalized integrated heats per injection as afunction of the molar ratio.

3.2. Structural and thermal stability

DSF experiments were performed on LZ both free in buffer andcomplexed with HA, to assess changes in the thermal stability of LZupon complexation. The DSF profile of LZ shows a sigmoidalincrease in fluorescence intensity followed by a decrease in fluo-rescence intensity as a function of temperature (Fig. 3). The initialincrease in fluorescence is most likely related to the increasedinteraction between the dye and the hydrophobic areas of theprotein that are exposed during thermal denaturation. The thermalunfolding temperature (Tm) was determined using the first

Fig. 3. DSF thermograms of LZ. (A) LZ in buffer. (B) LZ complexed with HA. (C) Firstderivative of LZ in buffer. (D) First derivative of LZ complexed with HA.

Table 1Tm for LZ and BSA determined by DSF (mean ± SD, n = 3).

Lysozyme Bovine serum albumin

0 mg/mL HA 0.768 mg/mL HA (molar ratio HA:LZ, 1:19.5) 0 mg/mL HA 0.768 mg/mL HA (molar ratio HA:BSA, 1:251)

0 mM NaCl 70.2 ± 0.3 70.6 ± 0.3 73.3 ± 0.8 72.4 ± 0.850 mM NaCl 70.4 ± 0.3 70.7 ± 0.3 74.6 ± 0.2 73.9 ± 0.3100 mM NaCl 69.6 ± 0.3 69.6 ± 0.3 74.3 ± 0.3 74.3 ± 0.6

Table 2Tm of LZ and complexes in 25 mM HEPES determined by DSC (n = 2).

Sample Tm (�C)

Lysozyme 74.2–74.3Complex coacervates 73.0–73.0

Fig. 4. Second derivative FTIR-spectra of LZ in buffer (dotted line) and complexes(solid line).

Table 3Area overlap of the second derivative area normalized FTIR spectra of free LZ and LZcomplexes in 25 mM HEPES buffer (means ± SD, n = 3, p < 0.0001 one-way ANOVA).

Samples Area overlap

Lysozyme vs. lysozyme 0.96 ± 0.002Complex vs. complex 0.98 ± 0.01Complex vs. lysozyme 0.95 ± 0.01

Table 4Relative recovery of BSA and LZ after 12 weeks of storage at room temperature(means ± SD, n = 3).

Sample Hyaluronic acid(mg/mL)

Ionic strength(mM)

Relativestability (%)

Lysozyme (2.4 mg/mL) 0.00 6 89 ± 20.35 6 94 ± 20.70 6 102 ± 50.00 100 89 ± 20.35 100 89 ± 30.70 100 87 ± 2

Bovine serum albumin(2.4 mg/mL)

0.00 6 106 ± 70.35 6 101 ± 10.70 6 100 ± 10.00 100 100 ± 10.35 100 98 ± 10.70 100 99 ± 4

J.J. Water et al. / European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331 329

derivative of the fluorescence intensity (Fig. 3). Based on the Tm

values (Table 1) the addition of sodium chloride has a small butsignificant effect (p < 0.001) on the thermal stability of LZ. Thusthe Tm decreases from 70.2 �C to 69.6 �C by adding 100 mM sodiumchloride. The effect of HA on Tm is not significant (p > 0.05). Theeffect of HA on the Tm of lysozyme was also determined usingDSC (Table 2, thermograms S1). This data showed a minor decreaseof the Tm of approximately 1 �C upon complexation of LZ with HA.

The thermal unfolding temperature for BSA was also deter-mined with the first derivative of the fluorescence intensity fromthe DSF thermograms (S2). For BSA the addition of sodium chlorideincreases the Tm value (p < 0.001) and HA has no effect (p > 0.05) onthe Tm value (Table 1).

Fig. 4 shows the second derivative FTIR spectra of LZ in bufferand the complexes. The structural changes were quantified bycomparing the area overlap of the different samples. The changein area overlap (Table 3) of the complexes compared to free LZwas minor, 5%, but nonetheless significantly different comparedto the variability of the control spectra (4% for LZ in buffer and2% for the complexes). However, the minor differences could notbe assigned to any specific changes in the secondary structure ofLZ upon complexation, as can be observed by visual inspection ofthe spectra, which showed no specific changes in peaks or values.

The stability study showed a significant improvement of the LZrecovery upon complexation with HA (Table 4) compared to free LZ(89 ± 2%). The effect was concentration dependent leading toincreased recovery rates 94 ± 2% and 102 ± 5% at respectively0.35 mg/mL (molar ratio HA:LZ, 1:8.9) and 0.70 mg/mL (molarratio HA:LZ, 1:17.7) HA. This could be explained by the incomplete

removal of LZ from the solution by complex coacervation at thelow HA ratio as shown by the solution depletion results. At highionic strength at which complex coacervation is inhibited no influ-ence on the physical stability of LZ was observed. The samples con-taining BSA showed no change in recovery upon storage with HA atthe different concentrations and ionic strengths tested.

4. Discussion and conclusion

The solution depletion and ITC experiments were able to char-acterize the complexation behavior of LZ with HA. The data of bothmethods were in agreement about the derived binding stoichiom-etry, providing a value of approximately 0.1 LZ molecules boundper HA monomer. This binding stoichiometry is in agreement withthe findings presented by Morfin et al. [22] of 0.125 per monomer,as obtained using Small Angle Neutron Scattering (SANS) in lowionic strength conditions, even though different chain length HA,buffer and mixing ratios were used. The ITC data also showed thatthe main driving force for complex coacervation was of entropicnature (DH = �0.3 kcal/mol and �TDS = �6.5 kcal/mol). Since theexperiments were performed in HEPES buffer (ionization enthalpy,4.9 kcal/mol) the observed binding enthalpy may contain a contri-bution from buffer de/protonation in case a net proton exchangecoupled to LZ binding to HA occurs. Therefore, some reservationmust be taken when interpreting the enthalpy and entropy values.

330 J.J. Water et al. / European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331

In addition, ITC experiments clearly showed a remarkable bindingcooperativity for LZ binding to HA. The dissociation constant for anisolated LZ molecule bound to HA is 13 lM (binding Gibbs energyof �6.8 kcal/mol); the binding of a neighbor LZ molecule on HAwould be accompanied by a 38-fold increase in binding affinity(cooperativity Gibbs energy of �2.2 kcal/mol), most likely promot-ing the formation of linear clusters of LZ molecules along the satu-ration process. This hypothesis is supported by the SANS datapresented by [22] describing the formation of a rod-like structureby binding of LZ to HA with lengths corresponding to the molecularlength of HA, which is not observed for complex coacervationbetween HA and pectin [27].

Ionic strength had a significant effect on the binding stoichiom-etry of the complexation as shown by the inhibition of complexcoacervation at ionic strengths exceeding 100 mM, although solu-ble complexes might still be present in solution [22]. At intermedi-ate ionic strength of 50 mM the presence of counter ions negativelyimpacted the complex coacervation behavior, resulting in a lowerbinding stoichiometry due to a less effective interaction betweenthe polymer and LZ (data not shown). From this data it can be con-cluded that the main driving force for the coacervation formationbetween HA and LZ was of electrostatic nature, since the presenceof counter ions was able to inhibit the complexation at physiolog-ical ionic strength (100 mM).

There was no effect on the thermal stability of LZ, as evidentfrom the fact that HA did not reduce the thermal induced denatur-ation of LZ upon complex coacervation as determined by DSF (Tm).This was emphasized by the non-significant change in Tm in case ofaddition of counterions that inhibited complexation. These resultswere not completely in agreement with the DSC data, whichshowed a minor, approximately 1 �C, reduction of the Tm uponcomplexation of the protein. This finding is in agreement with lit-erature that shows that anionic polymers with few hydrophobicgroups, like HA, usually do not significantly alter the protein struc-ture and therefore have a non-significant effect on the protein sta-bility [29]. This makes HA more suitable for protein formulationcompared to other polyanions that have been shown to reducethermal protein stability upon complexation [29,31].

As evident from the stability experiment, LZ stability was mostlikely increased with regards to the formation of insoluble aggre-gates by the formation of a complex coacervates with HA. This sta-bility improvement is presumably caused by the stronginteractions of the polyanion HA with the cationic charge patcheson the LZ protein surface, leading to the complex coacervationand reduction of the protein mobility but no change in LZ structure(FTIR) and unfolding temperature (DSF). We did not obtain conclu-sive data regarding the potential (de)stabilizing effect of HA onBSA, as the control samples without HA also showed a completerecovery of the BSA. Since the formation of soluble complex coac-ervates is described in the literature for both BSA and LZ [18], andno conclusive data for BSA were obtained, we cannot determinewhether the increased protein stability of LZ was only due to theformation of soluble complexes or due to the macroscopic complexcoacervates. However, the increased stability was mainly observedin samples showing macroscopic complexation.

In conclusion this study shows that HA is able to form a com-plex coacervate with lysozyme without affecting the protein sec-ondary structure or negatively impacting the thermal stability.Furthermore, the coacervation was able to prevent the formationof insoluble aggregates, which in turn led to increased storage sta-bility of the protein. In addition the complex coacervates were dis-sociated in 100 mM NaCl, which is below physiological ionicstrength and are therefore expected to dissociate in vivo uponadministration. This makes HA an interesting polymer for formula-tion development with regard to stability of the protein-basedactive pharmaceutical ingredient.

Acknowledgments

The authors acknowledge Novo Nordisk for partially fundingthe MicroCal VP-DSC, the Drug Research Academy for fundingthe NanoDrop 2000C, the Advanced Technology Foundation forfunding the TA Instruments Nano-ITC system and Apotekerfondenaf 1991 for funding the Bomem MB-100 FTIR spectrometer. Novo-zymes A/S is acknowledged for providing the hyaluronic acid usedfor this study and for financing a scholar grant.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ejpb.2014.09.001.

References

[1] S. Al-Qadi, M. Alatorre-Meda, E.M. Zaghloul, P. Taboada, C. Remunán-López,Chitosan–hyaluronic acid nanoparticles for gene silencing: the role ofhyaluronic acid on the nanoparticles’ formation and activity, Colloids Surf.,B: Biointerfaces 103 (2013) 615–623.

[2] M. Anderot, M. Nilsson, A. Végvári, E.H. Moeller, M. van de Weert, R. Isaksson,Determination of dissociation constants between polyelectrolytes and proteinsby affinity capillary electrophoresis, J. Chromatogr., B: Anal. Technol. Biomed.Life Sci. 877 (2009) 892–896.

[3] K. Aune, C. Tanford, Thermodynamics of the denaturation of lysozyme byguanidine hydrochloride. I. Dependence on pH at 25�*, Biochemistry (1969)4579–4585.

[4] E.A. Balazs, Viscosupplementation for treatment of osteoarthritis: from initialdiscovery to current status and results, Surg. Technol. Int. 12 (2004) 278–289.

[5] M. Brouwer, R. Cashon, J. Bonaventura, Functional properties of hemoglobinimmobilized in coacervates prepared from gelatin A and polyanioniccarbohydrates, Biotechnol. Bioeng. 35 (1990) 831–836.

[6] H.G. Bungenberg de Jong, Complex colloid systems, in: H.R. Kruyt (Ed.), ColloidScience, Elsevier Publishing Company, New York, 1949, pp. 335–432.

[7] K. Chung, J. Kim, B.-K. Cho, B.-J. Ko, B.-Y. Hwang, B.-G. Kim, How does dextransulfate prevent heat induced aggregation of protein? The mechanism and itslimitation as aggregation inhibitor, Biochim. Biophys. Acta 1774 (2007)249–257.

[8] T. Coviello, P. Matricardi, C. Marianecci, F. Alhaique, Polysaccharide hydrogelsfor modified release formulations, J. Controlled Release 119 (2007) 5–24.

[9] A. Dong, P. Huang, W.S. Caughey, Protein secondary structures in water fromsecond-derivative amide I infrared spectra, Biochemistry 29 (1990)3303–3308.

[10] X. Du, P.L. Dubin, D.A. Hoagland, L. Sun, Protein-selective coacervation withhyaluronic acid, Biomacromolecules 15 (2014) 726–734.

[11] E. He, C.Y. Yue, F. Simeon, L.H. Zhou, H.P. Too, K.C. Tam, Polyplex formationbetween four-arm poly(ethylene oxide)-b-poly(2-(diethylamino)ethylmethacrylate) and plasmid DNA in gene delivery, J. Biomed. Mater. Res. A 91(2009) 708–718.

[12] T. Hirakura, K. Yasugi, T. Nemoto, M. Sato, T. Shimoboji, Y. Aso, N. Morimoto, K.Akiyoshi, Hybrid hyaluronan hydrogel encapsulating nanogel as a proteinnanocarrier: new system for sustained delivery of protein with a chaperone-like function, J. Controlled Release 142 (2010) 483–489.

[13] M.R. Kasimova, A. Velazquez-Campoy, H.M. Nielsen, On the temperaturedependence of complex formation between chitosan and proteins,Biomacromolecules 12 (2011) 2534–2543.

[14] A.B. Kayitmazer, D. Seeman, B.B. Minsky, P.L. Dubin, Y. Xu, Protein–polyelectrolyte interactions, Soft Matter 9 (2013) 2553–2583.

[15] B.S. Kendrick, A. Dong, S.D. Allison, M.C. Manning, J.F. Carpenter, Quantitationof the area of overlap between second-derivative amide I infrared spectra todetermine the structural similarity of a protein in different states, J. Pharm. Sci.85 (1996) 155–158.

[16] W.-F. Lai, M.C.-M. Lin, Nucleic acid delivery with chitosan and its derivatives, J.Controlled Release 134 (2009) 158–168.

[17] S.B. Lee, Y.M. Lee, K.W. Song, M.H. Park, Preparation and properties ofpolyelectrolyte complex sponges composed of hyaluronic acid and chitosanand their biological behaviors, J. Appl. Polym. Sci. 90 (2003) 925–932.

[18] H. Lenormand, B. Deschrevel, F. Tranchepain, J.-C. Vincent, Electrostaticinteractions between hyaluronan and proteins at pH 4: how do theymodulate hyaluronidase activity, Biopolymers 89 (2008) 1088–1103.

[19] X. Liu, C. Liu, C.V. Catapano, L. Peng, J. Zhou, P. Rocchi, Structurally flexibletriethanolamine-core poly(amidoamine) dendrimers as effective nanovectorsto deliver RNAi-based therapeutics, Biotechnol. Adv. (2013) 1–9.

[20] O. Malay, O. Bayraktar, A. Batigün, Complex coacervation of silk fibroin andhyaluronic acid, Int. J. Biol. Macromol. 40 (2007) 387–393.

[21] K. Meyer, J. Palmer, The polysaccharide of the vitreous humor, J. Biol. Chem.(1934) 629–634.

[22] I. Morfin, E. Buhler, F. Cousin, I. Grillo, F. Boué, Rodlike complexes of apolyelectrolyte (hyaluronan) and a protein (lysozyme) observed by SANS,Biomacromolecules 12 (2011) 859–870.

J.J. Water et al. / European Journal of Pharmaceutics and Biopharmaceutics 88 (2014) 325–331 331

[23] W. Park, K. Na, Dermatan sulfate as a stabilizer for protein stability inpoly(lactide-co-glycolide) depot, Biotechnol. Bioprocess Eng. 14 (2009)668–674.

[24] L.E. Prevette, E.N. Nikolova, H.M. Al-Hashimi, M.M. Banaszak Holl, Intrinsicdynamics of DNA–polymer complexes: a mechanism for DNA release, Mol.Pharm. 9 (2012) 2743–2749.

[25] H. Ragelle, R. Riva, G. Vandermeulen, B. Naeye, V. Pourcelle, C.S. Le Duff, C.D’Haese, B. Nysten, K. Braeckmans, S. de Smedt, C. Jérôme, V. Préat, Chitosannanoparticles for siRNA delivery: optimizing formulation to increase stabilityand efficiency, J. Controlled Release 17 (2013) 54–63.

[26] K. Rawat, V.K. Aswal, H.B. Bohidar, DNA–gelatin complex coacervation, UCSTand first-order phase transition of coacervate to anisotropic ion gel in 1-methyl-3-octylimidazolium chloride ionic liquid solutions, J. Phys. Chem. B116 (2012) 14805–14816.

[27] I. Schmidt, F. Cousin, C. Huchon, F. Boué, M.A.V. Axelos, Spatial structure andcomposition of polysaccharide–protein complexes from small angle neutronscattering, Biomacromolecules 10 (2009) 1346–1357.

[28] J.E. Scott, F. Heatley, Hyaluronan forms specific stable tertiary structures inaqueous solution: a 13C NMR study, Proc. Natl. Acad. Sci. USA 96 (1999)4850–4855.

[29] E. Sedlák, D. Fedunová, V. Veselá, D. Sedláková, M. Antalík, Polyanionhydrophobicity and protein basicity affect protein stability in protein–polyanion complexes, Biomacromolecules 10 (2009) 2533–2538.

[30] S.L. Turgeon, C. Schmitt, C. Sanchez, Protein–polysaccharide complexes andcoacervates, Curr. Opin. Colloid Interface Sci. 12 (2007) 166–178.

[31] M. Van de Weert, M.B. Andersen, S. Frokjaer, Complex coacervation oflysozyme and heparin: complex characterization and protein stability,Pharm. Res. 21 (2004) 2354–2359.

[32] G.W. Vandenberg, C. Drolet, S.L. Scott, J. de la Noüe, Factors affecting proteinrelease from alginate–chitosan coacervate microcapsules during productionand gastric/intestinal simulation, J. Controlled Release 77 (2001) 297–307.

[33] A. Velazquez-Campoy, Ligand binding to one-dimensional lattice-likemacromolecules: analysis of the McGhee-von Hippel theory implemented inisothermal titration calorimetry, Anal. Biochem. 348 (2006) 94–104.

[34] K.P. Vercruysse, G.D. Prestwich, Hyaluronate derivatives in drug delivery, Crit.Rev. Ther. Drug Carrier Syst. 15 (1998) 513–555.

[35] I.A. Veselova, L.I. Malinina, P.V. Rodionov, T.N. Shekhovtsova, Properties andanalytical applications of the self-assembled complex {peroxidase–chitosan},Talanta 102 (2012) 101–109.